next up gif
Next: Visible vs. Infrared Up: Cirrus Cloud Visible Previous: Calibration Technique
Return to the Publications. Return to the Index.

Calibration Results

The results of the calibrated, attenuation corrected VIL data between the heights of 6 to 9 km above the HSRL are seen in Figure 12. The point comparison between the HSRL and the calibrated VIL shows more scatter than the comparison for the data between the heights of 6 and 7.5 km. This was a result of the misalignment errors between the two profiles described earlier and the usage of a bulk aerosol backscatter phase function for the correction of the attenuation in the VIL signal.

The lowest detected by the VIL at a horizontal distance of 24 km was approximately 3 10 m sr (-6.5 in Figure 12). The VIL data centered around a value of 5.0 10 m sr (-6.3) corresponded to non-cirrus aerosol backscatter. Little correlation was expected in this range because of the wavelength dependence of the scattering at the two lidar wavelengths by the non-cirrus aerosols. The centered around 1.0 10 msr (-5.0) correspond to the backscatter by cirrus cloud ice particles.

In Figure 12, the VIL calibrated are approximately 1.5 times greater than the HSRL . This was a result of the VIL signal being corrected for multiple scattering while the HSRL signal was not. If a multiple scattering correction factor of 0.5 was included in the calculation of the HSRL (similar to the VIL multiple scattering correction), then in Equation 10 would increase resulting in an increase in .

  
Figure 12: VIL--HSRL cirrus cloud data point comparison on December 1, 1989 from 19:29 to 21:20 GMT. The x-axis is the logarithm of the calibrated VIL aerosol backscatter cross sections. The y-axis is the logarithm of the HSRL aerosol backscatter cross sections. The point comparison is for data between 6 km and 9 km in height, the depth of the cirrus cloud layer. A one to one line is plotted for reference.

The calibration of the VIL data was extended to the cross wind scans for the two time periods. This extension produced aerosol backscatter cross sections for the mesoscale volume at a resolution of 2-3 km parallel to the wind (a function of the wind speed and the scanning rate of the VIL) and a resolution of 60 m in both the cross wind scan direction and the vertical. To determine whether the VIL calibrated for the mesoscale volume had a similar distribution to the VIL simulated RTI and HSRL RTI distributions, 50 point histograms of the data between the heights of 6 km and 11 km were computed. To create the VIL histograms, the molecular was calculated using Equation 11 (for 1064 nm) and subtracted from the VIL calibrated backscattered signal. This allowed for the VIL and HSRL distributions for the atmospheric aerosols to be compared.

  
Figure 13: 50 point histograms of between heights of 6 km and 11 km from 19:29 to 21:20 GMT on December 1, 1989. The x-axis is the logarithm of and the y-axis is the percentage of points falling within each interval. Histograms of the HSRL (short dash), vertical profile of VIL over the HSRL position (24 km to the East of the VIL (solid line)), and the VIL throughout a mesoscale volume produced from consecutive VIL cross wind scans (long dash) are shown. The cirrus cloud threshold value used in Section 3 coincides with a value of 1.0 10 m sr.

The resulting histograms are seen in Figure 13. A peak occurs in all three histograms near 8 10 msr (-5.2). This peak was associated with the cirrus ice particle . In all three histograms, the range of for the ice crystals spanned from 1 10 to 1 10 msr. The peak at the smaller , which represents the non-cirrus aerosols, occurred near 1.5 10 m sr for the HSRL RTI and the VIL data from the mesoscale volume. The non-cirrus aerosol peak for the VIL RTI, which occurred at 5.0 10 msr, shows the loss of signal at far ranges for the VIL. For below 3.8 10 msr (neglecting attenuation affects), the VIL receiver was not sensitive enough to detect the radiation backscattered from a horizontal distance of 24 km. The limit of detectability for a distance of 6 km can be seen in the histogram of the VIL data for the mesoscale volume. The smallest detectable was 1.25 10 msr (-6.8) as seen in Figure 13. Some HSRL were smaller than 1.0 10 msr. This was a result of incomplete separation between the aerosol and molecular channels for regions with small aerosol backscatter. This incomplete separation even led to some negative HSRL beneath the cirrus cloud layer. This problem has been removed in a new configuration of the HSRL (Piironen and Eloranta (1993)). Figure 13 also shows a relative minimum value for the near 1.0 10 msr for all three histograms. This relative minimum shows a clear separation between the background aerosol backscatter and the ice particle backscatter. This minimum value coincides with the threshold value used for the cirrus cloud determination in Section 3. At 1064 nm, the molecular backscatter cross section at a height of 6 km (calculated from the coincident rawinsonde density profile) was 6.44 10 msr. A typical aerosol backscatter cross section taken from the cirrus volume was 1.58 10 msr. By comparing these two values, it was determined that the signal backscattered by the background aerosols at a wavelength of 1064 m was at least a factor of twenty greater than the molecular backscatter at the cirrus cloud heights during this experiment.

Although the peak at the ice crystal backscatter cross sections coincides for the three histograms, large variations are seen in the calibration plot between the VIL and the HSRL (Figure 12). The scatter seen in Figure 12 can result from variations in the throughout the cirrus mesoscale volume, misalignments between the two lidar systems, and/or signal noise. The are dependent on the shape (or type) of the ice crystals scattering the laser light. The for different types of ice crystals can change by 0.048 sr as calculated by Takano and Liou (1989) (the values for the ice crystal varied between 0.037 and 0.085 sr). If the cirrus changed significantly over the volume, then the actual will change along with the causing errors in the attenuation correction technique. A second cause for the errors in the calibration may be a result of the variations of the ice crystal across the cirrus clouds.

The variation of the for visible wavelengths within the cirrus clouds will give an estimate on the allowable error between the VIL scan plane and the HSRL position. If the cirrus particles vary significantly from point to point then the alignment of the VIL has to be precise. If, on the other hand, the for the cirrus particles vary slowly within the cloud, then small alignment errors will be acceptable. To determine the variation of the ice crystal between the cirrus cloud data points in the scanned mesoscale volume, autocorrelations of the along wind and cross wind cirrus scans were calculated.

During the first time period (18:07-19:24 GMT) scattered cirrus clouds occurred throughout the mesoscale volume. Along the wind, there was 95.6% correlation between cirrus cloud for data points separated by 100 meters in the horizontal. For a 200 meter horizontal separation, there was an 89.5% correlation between cirrus cloud points. At 500 meters, there was a 73.9% correlation and at a distance of 1000 meters, the correlation dropped to 55.9%. In the vertical, for data points separated by 60, 120 and 240 meters, there was an 82.9%, 58.9%, and a 31.4% correlation between the ice crystal , respectively. The cirrus cloud correlations were also calculated for the VIL cross wind scans. For a 100 meter horizontal separation along the scan plane, there was a 94.6% correlation. At 200 meters, an 88.2%, at 500 meters, a 73.2% correlation, and at 1000 meters, a 54.6% correlation between the cirrus cloud . In the vertical, at a 60 meter interval, there was a 79.0% correlation, 120 meters, a 49.1% correlation, and at 240 meter separation, a correlation of 29.3% was seen.

For this time period, the horizontal and vertical correlations were similar for the along and cross wind VIL scans. Good correlation existed for points separated by 100 to 200 meters. When the length of the correlation was extended to 500 and 1000 meters, the variation between data points became significant. The high correlation at 100 and 200 meter separations in the horizontal suggest that small azimuth angle alignment errors between the two systems can be tolerated. If the misalignments between the two profiles was as large as 1000 meters, then the resulting calibration would be very poor. In the vertical, only a maximum error of 60 m could be tolerated when generating a calibration for the VIL. The vertical cirrus cloud correlations changed more rapidly than the horizontal correlations. This was due to the vertical wind shear within the cirrus cloud. For the first time period, the larger separation lengths had correlations less than 50% for the cirrus ice crystal . A calibration for the VIL could not be produced for larger separation lengths in the horizontal and the vertical due to the cirrus cloud variations.

For the second time period (19:29-21:20 GMT), the cirrus clouds were more spatially uniform due to the presence of a cirrus cloud deck. For the along wind scan, at 100, 200, 500, and 1000 meter separations, correlations of 94.8%, 89.3%, 78.6%, and 67.9% were seen, respectively. In the vertical, at 60, 120, and 240 meter data point separations, correlations of 84.2%, 64.4%, and 45.3% were seen, respectively. For the cross wind scan, at 100, 200, 500, and 1000 meter separation between points, correlations of 95.2%, 89.3%, 75.6%, and 59.4% between the cloud data points existed. In the vertical, for 60 meter data point separation, an 82.1% correlation was seen while at 120 meters a 59.0% correlation occurred. At a 240 meter separation in the vertical, a correlation of 36.1% was detected. During this time period, the correlation values along the wind were higher than those in the cross wind scans for the larger distances between data points. This was a result of the widespread cirrus cloud deck over most of the region at this time which had an aspect ratio of 9:1.

For the VIL RTI simulating the HSRL RTI during the first time period, the correlation between cloud points was calculated. The correlation was computed along the wind direction with a separation between points (scans) in the horizontal of approximately 3 km and with a 60 meter data point separation in the vertical. In the horizontal, correlations of 76.3%, 48.2%, 18.1%, and 15.5% were seen at 1, 2, 5, and 10 scan separations (points). In the vertical, correlations of 89.2%, 71.0%, and 52.7% were seen at 60, 120, and 240 meter separations, respectively. For the second time period, correlations of 83.3%, 65.6%, 46.9%, and 36.1% were seen for 1, 2, 5, and 10 scan separations respectively. In the vertical, separations of 60, 120, and 240 meters resulted in 88.9%, 73.3%, and 55.1% correlations between the cirrus cloud aerosol backscatter cross sections, respectively. Higher correlations were seen in the vertical because these points were taken from a single cirrus cloud scan while the horizontal data points were taken from consecutive scans. The vertical correlations were similar to the previously calculated values as would be expected. There was very little correlation between the cirrus cloud particulate in the horizontal because of the large distances separating consecutive data points and the 2 km averaging along the scans to produce each profile. The second time period had higher correlations in the horizontal than the first time period. This was a result of a widespread cirrus cloud deck which occurred during the second time period while the earlier period had scattered cirrus clouds throughout the region. The scattered cirrus clouds created a situation which made it difficult to produce a VIL RTI to match the HSRL RTI. The cirrus cloud deck on the other hand had more uniformity which allowed for greater scan angle errors between the two systems. Also, the VIL scan plane was more closely aligned with the wind direction at the cirrus cloud heights during the later period.

The cirrus cloud correlations revealed the types of error which would occur due to misalignments between the viewing positions of the two lidars. If large scan angle errors were present (greater than 1), then the cirrus cloud point comparison between the HSRL RTI and the VIL simulated RTI would be poor due to the variation of the within the cirrus clouds. For acceptable calibration results, the VIL azimuth angle errors have to be less than 0.5 and the elevation angle errors have to be less than 0.25.

For the two time periods, the cirrus cloud correlation values were similar but the calibration plots were not. The attempted calibration plot for the first time period (18:08 to 19:24 GMT) was very poor and as a result was not shown. The difference between the two time periods was the direction of the cirrus cloud advection and the widespread cirrus cloud deck throughout the later period. The clouds during the first time period were advected into the region from 283 while the clouds during the second time period came from 277 (on average). This difference of 6 between the wind direction and the VIL along wind scan direction result in point comparison errors in the calibration plot. The errors occurred when the VIL data was averaged along the scan plane to simulate the HSRL data (which was averaged along the wind direction). Since the VIL was averaged over the same distance as the HSRL profile (2 km), the offset between the VIL and HSRL data points at the end of the averaging length (1 km from each system) was 200 meters for the later time period and 300 meters for the early time period. The aerosol backscatter cross section correlation values across the wind dropped approximately 14% at the end of the averaging distances for the given time periods. This reduction in correlation, along with the scattered cirrus clouds, made the production of a VIL calibration plot from the first time period data impossible. This shows the importance of either aligning the VIL along the wind or using smaller averaging times in determining the cirrus cloud visible optical properties from the HSRL.



next up gif
Next: Visible vs. Infrared Up: Cirrus Cloud Visible Previous: Calibration Technique
Return to the Publications. Return to the Index.



Antti Piironen
Thu Apr 11 08:27:54 CDT 1996